Comparison of Quality of Life among Participants Enrolled in the Weight Watchers and Diabetes Prevention Programs, and Handling of Missing QOL Data: A Prospective Randomized Two-Arm Controlled Study

Abstract

Introduction: Weight Watchers (WW) is a commercial weight management company that offers various specialized programs and products to individuals seeking to lose and maintain weight. Diabetes Prevention Program (DPP) is evidence-based lifestyle change programs designed to prevent or delay type 2 diabetes. The DPP programs are available across the United States and are developed based on the partnership between public and private organizations. Both WW and DPP programs offer opportunities to individuals looking to lose weight, reduce the risk of type II diabetes, and improve quality of life (QoL). A prospective randomized two-arm controlled study enrolled 224 participants in the WW program (intervention arm) and DPP program (control arm) for a one year. The study was conducted in Indianapolis, Indiana, from 2013 to 2014. Data were collected at the baseline, after six months follow-up, and twelve months follow-up. Objectives: The study aimed: 1. To compare the QoL among individuals enrolled in the Weight Watchers and Diabetes Prevention programs. 2. To conduct secondary analyses assessing if the amount and rate of weight loss were associated with improved QoL. 3. To undertake an extensive sensitivity analysis in the handling of the QoL missing data. Methods: Randomized controlled study: The primary analysis compared QoL among participants enrolled in the Weight Watchers program (n = 110) and the National Diabetes Prevention program (n = 114) using a linear mixed model. The outcome of interest was a change in the QoL measured after 6 and 12 months of study enrollment. The secondary analysis assessed the relationship between the amount of weight loss and QoL using a linear model. We modeled the QoL at the end of the study as the response variable and the change in weight between baseline and 12 months as an explanatory variable. Another secondary analysis assessed if losing weight fast is associated with improved QoL. We defined fast weight loss as losing 5% or more of baseline weight by 6-month, slow weight loss as losing 5% or more of baseline weight by 12-month (excluding fast weight loss category), and no weight loss as losing less than 5% of initial weight throughout the study. We used a linear model to model the relationship between the QoL after the twelve months and indicator for fast, slow, and no weight loss categories. We adjusted for age, BMI, sex, and treatment group. QoL was an ordinal categorical variable with five levels (1 to 5), but in the modeling, we considered it to be a continuous variable with 1 indicating the poor QoL and 5 indicating the best QoL. Sensitivity analysis to the missing data: The QoL missing data were 8.3% (n=9) for intervention arm and 29.6% (n=29) for control arm at six-month follow-up, and were 13.5% (n=14) for intervention arm and 34.7% (n=34) for control arm at twelve-month follow-up. We performed in-depth sensitivity analysis using four different missing data imputation methods(multiple imputations by chained equations, random forest, k Nearest Neighbour, and MCMC). Sensitivity analysis to missing at random assumption: Multiple imputation models are useful for dealing with missing data, but they assume that the data are missing at random (MAR).We used tipping point analysis method to investigated the robustness of the MAR assumption-based results and conclusion. Results: The primary analysis: The average QoL (both arms combined) at baseline was 2.92, after six months it was 3.31, and at twelve months it was 3.29. The average QoL improvement from baseline to six month in the control arm was 0.37, 95%CI of 0.227 to 0.513, (p<.0001) and 0.39, 95%CI of 0.257 to 0.523, (p<.0001) for the intervention arm. After twelve-month, the average QoL improvement from baseline was 0.37, 95%CI of 0.223 to 0.517, (p<.0001) and 0.35, 95%CI of 0.214 to 0.494, (p<.0001) for the control and intervention arms respectively. The difference in change of QoL improvement between the control and intervention arms after six months was 0.17, 95%CI of -0.034 to 0.37, (p=0.59), and 0.14, 95%CI of -0.068 to 0.35, (p=0.81) after twelve months. We used a linear model to analyze the relationship between the weight change from baseline to the end of the study and the QoL, adjusting for baseline age, sex, BMI, and treatment group. The results showed that weight loss is significantly and positively related to the QoL. The QoL improved by 0.15, 95%CI of 0.07 to 0.24, (p<.0008) for every ten lbs of weight loss. Further, we used a linear model to analyze the relationship between QoL and weight loss of 5% or more within the first six months (fast weight loss) compared to slowly losing weight throughout the study (slow weight loss), and not losing at least 5% of weight (non-loser), adjusting for baseline age, sex, BMI, and treatment group. The results showed that losing at least 5% of weight within six months was not significantly better in improving QoL compared to losing weight slowly throughout the study. The QoL improved by 0.26, 95%CI of -0.21 to 0.73, (p=0.29), for a 5% or more of weight loss within six months. However, losing at least 5% of weight within six months was significantly related to improved QoL compared to non-loser. The QoL improved by 0.36, 95%CI of 0.08 to 0.65, (p=0.014), for a 5% or more of weight loss within six months. Sensitivity analysis: 1. The missing data: The results based on random forest, kNN, and MCMC imputation methods were very similar to the primary analysis results. However, the results based on MICE imputation method indicated that if we were to observe all the data, the difference in change of QoL improvement between the intervention and control arms(0.22, 95%CI of 0.03 to 0.41, p=0.027) would be significant after six months and not significant after twelve months of follow-ups(0.16, 95%CI of -0.04 to 0.37, p=0.124). 2. The missing at random assumption: The sensitivity analysis to MAR assumption and the primary analysis conclusions were similar. Therefore, the MAR assumption remained valid. Conclusions: Both Weight Watchers and Diabetes Prevention programs helped participants overall improve QoL, and the QoL improvement participants experienced from both programs were comparable. WeightWatchers were more effective in helping participants lose weight. Overall, sensitivity analysis did not change primary conclusions.